KGPFN: Unlocking the Potential of Knowledge Graph Foundation Model via In-Context Learning

2026-05-14

AI Summary

This work addresses the limitation of existing knowledge graph foundation models, which primarily focus on relation-level generalization and struggle to effectively integrate local and global structured context for zero-shot reasoning. The authors propose KGPFN, the first framework to systematically incorporate in-context learning into knowledge graph foundation models. It leverages relational graph message passing and multi-layer NBFNet to encode local neighborhoods, while constructing global context through relation instance retrieval. These components are unified within a Prior-Data Fitted Network framework, enabling strong generalization without fine-tuning. Evaluated across 57 benchmark datasets, KGPFN surpasses state-of-the-art models that require fine-tuning, demonstrating significantly enhanced transferability to unseen graphs solely through in-context learning.

Abstract

Knowledge graph (KG) foundation models aim to generalize across graphs with unseen entities and relations by learning transferable relational structure. However, most existing methods primarily emphasize relation-level universality, while in-context learning, the other pillar of foundation models remains under-explored for KG reasoning. In KGs, context is inherently structured and heterogeneous: effective prediction requires conditioning on the local context around the query entities as well as the global context that summarizes how a relation behaves across many instances. We propose KGPFN, a KG foundation model using Prior-data Fitted Network that unifies transferable relational regularities with inference-time in-context learning from structured context. KGPFN first learns relation representations via message passing on relation graphs to capture cross-graph relational invariances. For query-specific reasoning, it encodes local neighborhoods using a multi-layer NBFNet as local context. To enable ICL at global scale, it constructs relation-specific global context by retrieving a large set of instances of the query relation together with their local neighborhoods, and aggregates them within a Prior-Data Fitted Network framework that combines feature-level and sample-level attention. Through multi-graph pretraining on diverse KGs, KGPFN learns when to instantiate reusable patterns and when to override them using contextual evidence. Experiments on 57 KG benchmarks demonstrate that KGPFN achieves strong adaptation to previously unseen graphs through in-context learning alone, consistently outperforming competitive fine-tuned KG foundation models. Our code is available at https://github.com/HKUST-KnowComp/KGPFN.

Problem

• knowledge graph
• foundation model
• in-context learning
• relational reasoning
• structured context

Innovation

• In-Context Learning
• Knowledge Graph Foundation Model
• Relation Graph
• Prior-Data Fitted Network
• Structured Context